Adaptive hip exoskeleton assist control method and system

Abstract

The application provides a self-adaptive hip exoskeleton assistance control method and system, comprising: determining the motion starting state of a user; identifying the current motion mode of the user; if the personalized assistance strategy for the current motion mode has not been established, collecting the user's autonomous motion data in a non-assistance learning mode and generating the strategy; if the strategy has been established, evaluating the matching degree of the real-time motion state and the strategy; switching to a safety mode when the matching degree is lower than a safety threshold, otherwise generating an assistance control signal based on the strategy. The application constructs a personalized assistance model by learning the real motion data of the user itself, rather than relying on a general ideal curve, so that the assistance of the exoskeleton can be highly synchronized with the motion intention and habit of the user, significantly improving the accuracy, comfort and safety of the assistance, and realizing self-adaptive control for different users and different motion scenarios.

Description

Technical Field

[0001] This invention relates to the field of wearable robot technology, and more specifically, to an adaptive hip exoskeleton-assisted control method and system. Background Technology

[0002] Hip exoskeleton robots, as a typical wearable device, can provide effective movement assistance to users in scenarios such as rehabilitation training and assisted walking. The core lies in how the control system accurately and comfortably outputs assistance to match the user's movement intentions.

[0003] Currently, mainstream hip exoskeleton assistive control technologies typically use pre-set mathematical models to generate assistive trajectories. For example, many solutions utilize sine wave functions or oscillators based on central pattern generators (CPGs) to construct a curve simulating the angle changes of the human hip joint during the walking cycle. The control system then calculates and outputs the corresponding drive motor torque based on this pre-set trajectory, thereby driving the user's lower limb movement.

[0004] However, such methods based on idealized models have significant limitations in practical applications. First, mathematical curves such as sine waves are oversimplifications of complex human gait, and their shape, period, and phase inherently deviate from the movement characteristics of the real human body. This makes it difficult for the timing and magnitude of assistance provided by the exoskeleton to precisely match the user's actual needs, often resulting in the user feeling "dragged along by the machine" or "fighting against the machine," leading to a poor human-computer interaction experience.

[0005] More importantly, there are significant individual differences in height, weight, muscle strength, and gait habits among individuals. Even the gait of the same user can change under different fatigue states or different exercise scenarios (such as flat ground, uphill, and downhill). Existing technologies using fixed or universal trajectory models lack the ability to adapt to such individual differences and dynamic changes. Although some systems support manual parameter adjustment, the process is cumbersome and relies on professionals, making it difficult to achieve truly personalized and adaptive control.

[0006] Reference 1, patent document CN121468478A, discloses a hip exoskeleton and a walking assistance control method. It includes three active drive units: a sagittal plane drive unit for hip flexion / extension; a coronal plane drive unit for hip adduction / abduction; and a horizontal plane drive unit for hip internal / external rotation. These units are connected via a remote rotation center mechanism, allowing the horizontal plane active drive unit to be aligned with the hip joint rotation center via a virtual axis. However, this patent's technical solution still lacks the feature of personalized adjustment.

[0007] Reference 2, Xu Linghui, Yang Wei, Yang Canjun, et al. Real-time assistive control of hip exoskeleton based on motion prediction [J]. Robotics, 2021, 43 (4): 473-483. This reference discloses a real-time assistive control scheme for hip exoskeleton based on motion prediction. It realizes dynamic assistive regulation of hip exoskeleton through motion prediction strategy, and can complete assisted walking control according to the limb movement law, providing a technical reference for the motion control design of lower limb rehabilitation and assistive exoskeleton. However, the technical solution in this reference still has the defect of lacking personalized adjustment.

[0008] Reference 3, Liu Yu, Huang Yan, Zhou Zhihao, et al. Kinematic analysis of human walking with a hip-assisted exoskeleton [J]. Journal of Peking University (Natural Science Edition), 2024(03):422-430. This reference discloses a method for analyzing the kinematic characteristics of human walking. Through experimental analysis of the joint motion patterns and gait characteristics of human walking with an exoskeleton, it clarifies the coupling relationship between the hip-assisted exoskeleton and human-machine walking motion, providing a theoretical basis for the structural optimization and control strategy design of the hip-assisted exoskeleton. However, the technical solution in this reference still lacks the defect of personalized adjustment.

[0009] Therefore, existing technologies generally suffer from poor assistance effects and poor user experience due to model mismatch and lack of personalized adjustment capabilities, which urgently require a control method that can adapt to individual user characteristics to solve these problems. Summary of the Invention

[0010] In view of the deficiencies in the prior art, the purpose of this invention is to provide an adaptive hip exoskeleton-assisted control method and system.

[0011] An adaptive hip exoskeleton assistive control method according to the present invention includes the following steps: Determine the user's initial state of motion, and after determining that the user is in the initial state of motion, identify the user's current motion mode; Determine whether a personalized assistance strategy for the current motion mode has been established; If the personalized assistance strategy is not established, then in the preset non-assisted learning mode, the user's autonomous movement data is collected, and the personalized assistance strategy is generated based on the autonomous movement data. If the personalized assistance strategy has been established, the user's current real-time motion status is obtained, and the degree of matching between the real-time motion status and the personalized assistance strategy is evaluated. When the matching degree is less than a preset safety threshold, the assist output of the hip exoskeleton will be switched to a preset safety mode. When the matching degree is greater than or equal to the preset safety threshold, an assistance control signal is generated based on the personalized assistance strategy.

[0012] Preferably, the degree of matching between the real-time motion state and the personalized assistance strategy is calculated using the following formula: ; Wherein, P is the matching degree, which measures the matching degree between the real-time motion state and the personalized assistance strategy, and takes the value [0, 1]; k is a custom allocation coefficient, and its value range is (0, 1). The motor reference angle value at one point on the reference trajectory given by the personalized assistance strategy; To and The corresponding real-time measured value of the motor angle returned by the current encoder; The reference angular velocity value of one point on the reference trajectory given by the personalized assistance strategy; To and The corresponding real-time measurement value of angular velocity returned by the current encoder; This is the maximum weighted distance value calculated when generating the reference trajectory.

[0013] Preferably, at least one state parameter capable of characterizing the user's lower limb movement state is acquired; The state parameters are kinematic or physiological parameters, and the user's initial state of movement is determined by the state parameters.

[0014] Preferably, the state parameters include at least one selected from hip joint angle, hip joint angular velocity, limb acceleration, plantar pressure signal, and electromyographic signal.

[0015] Preferably, the step of identifying the user's current motion mode includes: The landing event is determined by the plantar pressure signal of the swing leg, and the running mode is determined by combining the pressure signal of the supporting leg. The mode of going upstairs, going downstairs, or walking on flat ground is determined by the hip joint angle at the moment of the landing event.

[0016] Preferably, the preset non-assisted learning mode is a zero-assistance mode.

[0017] Preferably, the personalized assistance strategy is a motion reference curve; The step of generating the personalized assistance strategy includes: processing the autonomous motion data to fit and generate the motion reference curve; The processing of the autonomous motion data includes filtering or smoothing the autonomous motion data.

[0018] Preferably, when it is determined that a personalized assist strategy for the current motion mode has not been established, the current autonomous motion data is recorded in chronological order, and the autonomous motion data includes angle data, angular velocity data, and torque value data; When it is determined that the recording of the autonomous motion data is complete, the recorded autonomous motion data is subjected to various filtering processes to obtain multiple corresponding reference motion curves; the various filtering processes adopt any of the following filtering methods: Kalman filtering, moving average filtering, and first-order hysteresis filtering; The weighted distance between the autonomous motion data and the corresponding filtered data is calculated point by point to obtain the maximum value of the weighted distance. The filtered data sequence corresponding to the minimum value is selected from the results of multiple filtering methods as the motion reference curve.

[0019] Preferably, within a preset number of cycles, if the number of times the matching degree between the real-time motion state and the personalized assistance strategy is less than a preset safety threshold exceeds a preset value, the motion reference curve is automatically cleared to zero, and new motion reference curves are generated by re-collecting data.

[0020] Preferably, the determination of the preset safety threshold includes setting it based on the maximum value of the weighted distance between the autonomous motion data and the generated personalized assistance strategy; The formula for calculating the maximum value of the weighted distance is: ; in, The maximum value of the calculated weighted distance; k is a user-defined allocation coefficient, with a value range of (0, 1); The motor reference angle value at one point on the reference trajectory given by the personalized assistance strategy; To and The corresponding motor angle measurement value obtained from recording autonomous motion data; The personalized assistance strategy provides the reference angular velocity value of one point on the reference trajectory; To and The corresponding motor angular velocity measurement value obtained from recording autonomous motion data; The preset safety mode sets the assist output torque of the hip exoskeleton to zero.

[0021] Preferably, the assist control signal is a torque control signal, which is sent to a motor for driving the hip exoskeleton; The formula for calculating the torque control signal is as follows: ; in, This is the torque control signal that is ultimately output to the motor; , , The coefficients are user-defined and can take values ​​from the entire real number field. , , The angle, angular velocity, and angular acceleration values ​​corresponding to the point on the motion reference curve that best matches the current real-time state in the proposed personalized assistance strategy. and This refers to the motor angle and angular velocity currently being acquired in real time.

[0022] Preferably, the step of identifying the user's current motion mode specifically includes the following steps: Step S1: Determine if a swing leg marker already exists. If yes, proceed to step S2. If no, calculate the changes in the hip joint angles of the left and right legs and mark the leg with a positive hip joint angle change as the swing leg. Step S2: Continuously calculate and monitor the plantar pressure value of the swing leg, and determine whether the plantar pressure value of the swing leg increases and exceeds the first threshold. If yes, proceed to step S3; otherwise, continue the calculation and monitoring. Step S3: Calculate the plantar pressure value of the supporting leg at the moment the swing leg lands, and determine whether the plantar pressure value of the supporting leg is less than the second threshold. If yes, mark the current state as running; otherwise, proceed to step S4. Step S4: Obtain the hip joint angles on both sides at the moment the swing leg lands, and determine whether the hip joint angle of the swing leg is greater than the third threshold. If yes, mark the current state as going upstairs; otherwise, proceed to step S5. Step S5: Determine whether the hip joint angle of the supporting leg is greater than the fourth threshold. If yes, mark the current state as walking on flat ground; otherwise, mark the current state as going downstairs.

[0023] The present invention also provides a hip joint exoskeleton assistive control system, which performs the following process: Determine the user's initial state of motion, and after determining that the user is in the initial state of motion, identify the user's current motion mode; Determine whether a personalized assistance strategy for the current motion mode has been established; If the personalized assistance strategy is not established, then in the preset non-assisted learning mode, the user's autonomous movement data is collected, and the personalized assistance strategy is generated based on the autonomous movement data. If the personalized assistance strategy has been established, the user's current real-time motion status is obtained, and the degree of matching between the real-time motion status and the personalized assistance strategy is evaluated. When the matching degree is less than a preset safety threshold, the assist output of the hip exoskeleton will be switched to a preset safety mode. When the matching degree is greater than or equal to the preset safety threshold, an assistance control signal is generated based on the personalized assistance strategy.

[0024] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention collects user's autonomous movement data in non-assisted learning mode and generates personalized assistance strategies based on this data, completely eliminating the reliance on general and idealized mathematical models. The assistance benchmark generated by this method is entirely derived from the user's own movement characteristics, enabling the exoskeleton's assistance to be highly synchronized with the user's movement intentions and habits, significantly improving the accuracy, following, and comfort of the assistance, and enhancing the human-computer interaction experience.

[0025] 2. This invention can identify the user's current movement mode, such as walking on flat ground, running, climbing stairs, etc., and establish or call up corresponding personalized assistance strategies for different modes. This enables the exoskeleton to intelligently provide the most suitable assistance mode when the user switches between different sports tasks, which significantly improves the adaptability and assistance efficiency of the exoskeleton in different life scenarios.

[0026] 3. This invention assesses the matching degree between the user's movement status and the established personalized assistance strategy in real time, and sets a safety threshold verification mechanism. When abnormal movements such as staggering or falling are detected or the system identification error is detected, the assistance output can be switched to the preset safety mode in a timely manner (for example, the assistance output torque is set to zero), thereby actively avoiding potential risks and greatly improving the safety of wearable use.

[0027] 4. This invention enables adaptive generation and adjustment of control strategies, avoiding the cumbersome manual parameter setting and debugging process in the prior art, lowering the threshold for using the equipment, enhancing the intelligence and automation level of the system, and enabling users to use the exoskeleton device more conveniently and efficiently. Attached Figure Description

[0028] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 A flowchart illustrating an adaptive hip exoskeleton-assisted control method; Figure 2 A flowchart illustrating the motion mode recognition process; Figure 3 This is a schematic diagram illustrating the fitting principle of the motion reference curve. Detailed Implementation

[0029] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0030] Example 1 This embodiment provides an adaptive hip exoskeleton assistive control method. This technical solution aims to solve the problems of rigid assistive control strategies in the prior art, which cannot be personalized to individual user differences and different sports scenarios.

[0031] This embodiment learns from the user's autonomous movement data in an uninterrupted state online, generates a unique personalized assistance strategy for each movement mode, and monitors the degree of matching between the user's movement state and the strategy in real time while providing assistance, thereby achieving accurate and safe adaptive assistance control.

[0032] In this embodiment, a "hip exoskeleton" refers to an electromechanical device worn on the waist and thigh that provides auxiliary or resistive torque at the hip joint via actuators such as motors. A "personalized assistance strategy" refers to a control benchmark or model generated for a specific user and specific motion modality to guide the assistance output. This can take the form of a parameterized motion reference curve, lookup table, or neural network model, reflecting the user's natural movement patterns in that modality. A "motion modality" refers to the type of activity performed by the user with significantly different kinematic or dynamic characteristics, such as walking on flat ground, climbing stairs, descending stairs, or running. Furthermore, a "non-assisted learning mode" refers to an exoskeleton operating mode where the system provides little or no assistance. Its purpose is to collect data that reflects the user's true, autonomous movement characteristics, avoiding data contamination caused by exoskeleton intervention. A "preset safety threshold" is a critical value used to determine whether the user's current movement state deviates from their normal pattern. This threshold can be adaptively set based on learned user autonomous movement data to improve the rationality and sensitivity of safety monitoring. Through the above design, this embodiment can realize closed-loop control of the entire process from user data collection and personalized strategy generation to safety assistance application on exoskeleton devices.

[0033] The method in this embodiment first needs to determine the user's initial movement state to determine whether the assist control logic needs to be activated. After determining that the user is in the initial movement state, the system will then identify the user's current movement mode. Subsequently, the system will check whether a corresponding personalized assist strategy has been established for the user and the movement mode. If not, the system will switch to a preset non-assist learning mode, in which mode it will collect the user's autonomous movement data and generate a new personalized assist strategy based on this data. If a strategy already exists, the system will enter the assist application stage, obtain the user's current real-time movement state, and evaluate the degree of matching between the state and the established personalized assist strategy.

[0034] The degree of matching is calculated using formula (1): (1); In formula (1), P refers to the degree of matching, which is used to measure the degree of matching between the real-time motion state and the personalized assistance strategy, and takes the value [0, 1]; k is a custom allocation coefficient, which takes the value range of (0, 1). In a specific embodiment, k can be taken as 0.5. It is the motor reference angle value corresponding to a certain point on the reference trajectory given by the personalized assistance strategy. It is the real-time measured value of the motor angle returned by the current encoder; It is the reference angular velocity corresponding to the reference trajectory point. It is the real-time measured value of the angular velocity returned by the current encoder.

[0035] The maximum weighted distance value calculated when generating the reference trajectory is specifically calculated using the following formula (2). The numerator of the second term in formula (1) refers to the minimum value that meets the calculation method by traversing all points on the motion reference curve corresponding to the current motion mode.

[0036] The assessment of this matching degree is a crucial step in ensuring safety. When the assessment result shows that the matching degree is below a preset safety threshold, it means that the user's current movement may be abnormal or deviating from the normal pattern. At this time, the system will switch the hip exoskeleton's assist output to a preset safety mode, such as stopping the application of assistance. Conversely, when the matching degree meets the safety requirements, the system will generate corresponding assist control signals based on the personalized assist strategy, driving the exoskeleton to provide assistance coordinated with the user's intentions. This series of steps constitutes a complete adaptive control closed loop, enabling the exoskeleton to gradually learn and adapt to the user's unique movement habits from "nothing" onwards, thereby providing a highly personalized and safe assist experience.

[0037] Building upon the aforementioned approach, to enable the system to perceive the user's motion state, this method includes a step of acquiring at least one kinematic or physiological parameter characterizing the user's lower limb motion state before determining the user's initial motion state. These parameters form the data foundation for all subsequent judgment, recognition, and control algorithms. By introducing sensors to collect these parameters, the control system can obtain real-time input about the user's body state from the physical world, which is a prerequisite for achieving intelligent and adaptive control.

[0038] Furthermore, as a preferred embodiment, the kinematic or physiological parameters may specifically include at least one selected from hip joint angle, hip joint angular velocity, limb acceleration, plantar pressure, and electromyography (EMG) signals. Hip joint angle and angular velocity directly reflect the hip joint's motion state and are core variables for generating assist and assessing fit. Limb acceleration, typically provided by an inertial measurement unit (IMU), can be used to assist in determining the start and stop of movement and posture changes. Plantar pressure can accurately capture key events in the gait cycle, such as heel strike and toe lift, and is crucial for motion mode recognition. EMG signals directly reflect the activation level of related muscles and can be used for deeper-level motion intention recognition. Using these multi-dimensional parameter combinations can significantly improve the accuracy and robustness of the system's user state assessment.

[0039] As a specific implementation method, the step of identifying the user's current motion modality can be based on a set of efficient rules designed according to the aforementioned parameters. For example, this step may include at least one step: using the plantar pressure signal of the swinging leg to determine the landing event, and combining it with the pressure signal of the supporting leg to determine the running modality; or, combining the hip joint angle at the moment of the landing event to determine the upstairs, downstairs, or flat-ground walking modality. This judgment logic based on specific biomechanical events and associated parameters has the advantages of low computational cost, fast response speed, and strong interpretability compared to complex machine learning models, making it very suitable for implementation on resource-constrained embedded control systems. This design provides a reliable, low-cost, and efficient specific implementation method for motion modality recognition.

[0040] To ensure that the generated personalized assistance strategy accurately reflects the user's natural movement habits, the preset non-assistance mode is preferably a zero-assistance mode. In this mode, the control system sends zero-torque or zero-current commands to the exoskeleton's actuators, allowing the motors to freely follow the user's movements, overcoming only their own friction and inertia without providing any active assistance or resistance. Therefore, the autonomous motion data collected by the system represents the user's pure movement performance under minimal external interference. The personalized assistance strategy generated based on this data thus has the highest fidelity, laying a solid foundation for subsequent precise assistance.

[0041] In a preferred embodiment, the personalized assistance strategy is specifically represented as a motion reference curve. Accordingly, the step of generating the personalized assistance strategy includes processing the collected autonomous motion data to fit and generate the motion reference curve. Concretizing the abstract strategy into a computable curve makes the assistance benchmark explicit and quantifiable. For example, the curve can describe the changes in hip joint angle, angular velocity, and even angular acceleration over time (or gait percentage) within a standardized gait cycle. This approach is not only intuitive but also facilitates subsequent interpolation calculations and deviation assessments, providing a quantifiable and computable personalized assistance benchmark.

[0042] To improve the quality of the motion reference curve, the processing of autonomous motion data may include filtering or smoothing the data. Raw sensor data often contains high-frequency noise and minor jitter during motion. Directly using this data for fitting can result in an unsmooth reference curve, or even unreasonable abrupt changes. Applying algorithms such as Kalman filtering, moving average filtering, or spline smoothing can effectively remove this noise and jagged edges, extracting the core trend of the motion. This processing makes the final fitted motion reference curve smoother and more stable, and the assist control signal generated based on this curve will also be smoother, thereby improving user comfort and the fluency of human-computer interaction.

[0043] Regarding the fitting and generation of the motion reference curve, when the algorithm determines that there is no corresponding motion reference curve, it enters the data recording process and records the current autonomous motion data in chronological order, that is, storing the angle, angular velocity and torque values ​​into the structure array. When the algorithm determines that the recording is complete, it performs various filtering processes on the recorded autonomous motion-related data, including but not limited to Kalman filtering, moving average, first-order hysteresis filtering and other filtering methods, so as to obtain multiple corresponding reference motion curves. Then, according to the following formula (2), the weighted distance between the autonomous motion data and the corresponding filtered data is calculated point by point to obtain the maximum value of the weighted distance. The minimum value corresponding to the minimum value is selected from the results of multiple filtering methods, and the filtered data sequence is used as the motion reference curve. The specific implementation process is as follows: Figure 3 As shown.

[0044] There are three ways to update the motion reference curve: a. The motion reference curve is automatically cleared or saved after each power-on or power-off cycle; the user can decide whether to switch functions. b. Users can manually clear the motion curve; c. If, under the current motion mode, the matching degree falls below the preset safety threshold multiple times within several cycles, the motion reference curve will be automatically cleared, and new motion reference curves will be generated by re-collecting data.

[0045] Regarding the safety design, the determination of the preset safety threshold can be linked to the user's individual movement characteristics. Specifically, the determination process may include setting it based on the maximum value of the weighted distance between the collected autonomous movement data and the finally generated personalized assistance strategy. For example, after generating the reference curve, the deviation (such as the weighted distance) between each original data point and the corresponding phase point on the curve can be calculated, and the maximum value among these deviations can be found. Then, the preset safety threshold is set to a multiple of this maximum value (e.g., 1.2 to 1.5 times). This adaptive threshold setting method makes the safety boundary no longer a fixed, universal value, but rather matches the fluctuation range of the user's own movement. For users with good movement stability, the threshold will be smaller, and the monitoring will be more sensitive; for users with large movement fluctuations, the threshold will be larger, and the tolerance will be higher. This design greatly improves the rationality of safety judgment and avoids false alarms or missed alarms caused by improper threshold settings.

[0046] The formula for calculating the maximum weighted distance is as follows: (2); in, The maximum value of the weighted distance is calculated, and k is a user-defined allocation coefficient with a range of (0, 1), consistent with the definition and value in formula (1). It is the motor reference angle value corresponding to a certain point on the pre-selected reference trajectory obtained after filtering. It is the motor angle measurement value obtained by recording autonomous motion data; It is the reference angular velocity corresponding to the pre-selected reference trajectory point. It is the measured value of the motor's angular velocity obtained by recording autonomous motion data.

[0047] When the system determines that the user's movement is abnormal and triggers the safety mechanism, the preset safety mode preferably sets the assist output torque of the hip exoskeleton to zero. This is the most direct and inherently safe protection measure. Once the deviation between the user's real-time movement and the personalized strategy exceeds the adaptively set safety threshold, the controller immediately abandons the calculation and output of assist torque, setting the command sent to the motor to zero. This means that the exoskeleton will immediately stop applying any active force to the user and switch to a passive following state. This design ensures that in any uncertain or potentially dangerous situation (such as the user tripping, falling, or making unexpected movements), the exoskeleton will not provide incorrect assistance that may exacerbate the danger, thereby maximizing the user's personal safety.

[0048] Ultimately, the control strategy generated at the algorithmic level needs to be translated into physical forces. Therefore, the assist control signal is specifically a torque control signal, and the method further includes sending this torque control signal to a motor that drives the hip exoskeleton. This torque control signal is the final output of the entire adaptive control process; it carries a precise torque value calculated based on the personalized assist strategy to assist the user in completing the current action. Through a digital-to-analog converter (DAC) and a motor driver, this signal is translated into a corresponding current, driving the motor to generate the target torque, which acts on the user's hip joint. This step connects the intelligent decision-making at the algorithmic level with the power output at the physical execution level, completing the entire control chain from perception and decision-making to execution.

[0049] The formula for calculating the torque control signal is as follows: (3); in, This is the final torque control signal output to the motor. , and For custom coefficients, the value can be taken from the entire real number field. In a specific embodiment, it can be taken as... , , ; , and In the proposed personalized assistance strategy, the angle, angular velocity, and angular acceleration values ​​corresponding to the point on the motion curve that best matches the current real-time state are used. and This refers to the motor angle and angular velocity currently being acquired in real time.

[0050] This embodiment also provides a hip exoskeleton assistive control device, which includes a processor and a memory. When the computer program stored in the memory is executed by the processor, it can implement the methods described in any of the above embodiments. This embodiment also provides a computer-readable storage medium, on which a computer program stored, when executed by a processor, can similarly implement the methods described in any of the above embodiments.

[0051] The present invention also provides an adaptive hip exoskeleton assistive control system, which can be implemented by executing the process steps of the adaptive hip exoskeleton assistive control method. That is, those skilled in the art can understand the adaptive hip exoskeleton assistive control method as a preferred embodiment of the adaptive hip exoskeleton assistive control system.

[0052] Example 2 This embodiment provides an adaptive hip exoskeleton assistive control system, which executes the following process: Determine the user's initial state of motion; after determining that the user is in the initial state of motion, identify the user's current motion mode; determine whether a personalized assistance strategy for the current motion mode has been established; If the personalized assistance strategy is not established, the user's autonomous movement data is collected in the preset non-assisted learning mode, and the personalized assistance strategy is generated based on the autonomous movement data; the preset non-assisted learning mode is a zero-assistance mode; if the personalized assistance strategy is established, the user's current real-time movement state is obtained, and the degree of matching between the real-time movement state and the personalized assistance strategy is evaluated. When the matching degree is less than a preset safety threshold, the assist output of the hip exoskeleton is switched to a preset safety mode; when the matching degree is greater than or equal to the preset safety threshold, an assist control signal is generated based on the personalized assist strategy.

[0053] Furthermore, the degree of matching between the real-time motion state and the personalized assistance strategy is calculated using the following formula: ; Wherein, P is the matching degree, which measures the matching degree between the real-time motion state and the personalized assistance strategy, and takes the value [0, 1]; k is a custom allocation coefficient, and its value range is (0, 1). The motor reference angle value at one point on the reference trajectory given by the personalized assistance strategy; To and The corresponding real-time measured value of the motor angle returned by the current encoder; The reference angular velocity value of one point on the reference trajectory given by the personalized assistance strategy; To and The corresponding real-time measurement value of angular velocity returned by the current encoder; This is the maximum weighted distance value calculated when generating the reference trajectory.

[0054] Furthermore, the module for identifying the user's current motion mode specifically includes the following modules: Module M1: Determine if a swing leg identifier already exists. If yes, execute module M2. If no, calculate the changes in the hip joint angles of the left and right legs and identify the leg with a positive hip joint angle change as the swing leg. Module M2: Continuously calculates and monitors the plantar pressure value of the swing leg, and determines whether the plantar pressure value of the swing leg has increased and exceeded the first threshold. If so, execute module M3; otherwise, continue to perform calculation and monitoring. Module M3: Calculates the plantar pressure value of the supporting leg at the moment the swing leg lands, and determines whether the plantar pressure value of the supporting leg is less than the second threshold. If yes, the current state is marked as running; otherwise, module M4 is executed. Module M4: Obtain the hip joint angles on both sides at the moment the swing leg lands, and determine whether the hip joint angle of the swing leg is greater than the third threshold. If yes, mark the current state as going upstairs; otherwise, execute module M5. Module M5: Determines whether the hip joint angle of the supporting leg is greater than the fourth threshold. If yes, it marks the current state as walking on flat ground; otherwise, it marks the current state as going downstairs.

[0055] Furthermore, at least one state parameter capable of characterizing the user's lower limb movement state is acquired; the state parameter is a kinematic parameter or a physiological parameter, and the user's initial movement state is determined by the state parameter. The state parameter includes at least one selected from hip joint angle, hip joint angular velocity, limb acceleration, plantar pressure signal, and electromyographic signal.

[0056] Furthermore, the module for identifying the user's current movement mode includes: using the plantar pressure signal of the swinging leg to determine the landing event, and combining it with the pressure signal of the supporting leg to determine the running mode; and combining the hip joint angle at the moment the landing event occurs to determine the stair-climbing mode, stair-climbing mode, or flat-ground walking mode.

[0057] Furthermore, the personalized assistance strategy is a motion reference curve; the module for generating the personalized assistance strategy includes: processing the autonomous motion data to fit and generate the motion reference curve; the processing of the autonomous motion data includes filtering or smoothing the autonomous motion data.

[0058] Furthermore, when it is determined that a personalized assistance strategy for the current motion mode has not been established, the current autonomous motion data is recorded in chronological order. The autonomous motion data includes angle data, angular velocity data, and torque value data. When it is determined that the recording of the autonomous motion data is complete, the recorded autonomous motion data is subjected to various filtering processes to obtain multiple corresponding reference motion curves. The various filtering processes employ any of the following filtering methods: Kalman filtering, moving average filtering, and first-order hysteresis filtering. The weighted distance between the autonomous motion data and the corresponding filtered data is calculated point by point to obtain the maximum value of the weighted distance. The filtered data sequence corresponding to the minimum value is selected from the results of multiple filtering methods as the motion reference curve.

[0059] Furthermore, within a preset number of cycles, if the number of times the matching degree between the real-time motion state and the personalized assistance strategy is less than a preset safety threshold exceeds a preset value, the motion reference curve is automatically cleared to zero, and data is re-collected to generate a new motion reference curve.

[0060] Furthermore, determining the preset safety threshold includes setting the maximum value of the weighted distance between the autonomous motion data and the generated personalized assistance strategy; the preset safety mode is to set the assist output torque of the hip exoskeleton to zero. The formula for calculating the maximum value of the weighted distance is: ; in, The maximum value of the calculated weighted distance; k is a user-defined allocation coefficient, with a value range of (0, 1); The motor reference angle value at one point on the reference trajectory given by the personalized assistance strategy; To and The corresponding motor angle measurement value obtained from recording autonomous motion data; The personalized assistance strategy provides the reference angular velocity value of one point on the reference trajectory; To and The corresponding motor angular velocity measurement value obtained from recording autonomous motion data.

[0061] Furthermore, the assist control signal is a torque control signal, which is sent to the motor used to drive the hip joint exoskeleton. The calculation formula for the torque control signal is: ; in, This is the torque control signal that is ultimately output to the motor; , , The coefficients are user-defined and can take values ​​from the entire real number field. , , The angle, angular velocity, and angular acceleration values ​​corresponding to the point on the motion reference curve that best matches the current real-time state in the proposed personalized assistance strategy. and This refers to the motor angle and angular velocity currently being acquired in real time.

[0062] Example 3 Those skilled in the art can understand this embodiment as a more specific description of Embodiment 1 and Embodiment 2.

[0063] Please see Figure 1 This illustrates the flow of an adaptive assist control method in a basic embodiment of this example. This method can be executed by a control device integrated into a hip exoskeleton. In a typical application scenario, a new user is wearing the exoskeleton device for the first time. When the user is ready to begin movement, the control system first executes step S101, continuously acquiring various kinematic or physiological parameters characterizing the user's movement state. Then, in step S102, the system analyzes these parameters to determine the user's initial movement state. For example, when a continuous change in angular velocity is detected in the user's hip joint, the system determines that the user has begun movement.

[0064] After determining that the user is in the initial stage of movement, the process proceeds to step S103, where the system identifies the user's current movement mode. For example, by analyzing the user's cadence, stride length, and other characteristics, the system identifies the current mode as "walking on flat ground." Next, in step S104, the system queries its internal memory to determine whether a personalized assistance strategy for the user and the "walking on flat ground" mode has been established. Since this is the first time using the system, the query result is negative.

[0065] Therefore, the process jumps to step S106, where the system enters a preset non-assisted or low-intervention learning mode. In this mode, the exoskeleton does not provide active assistance, but only passively follows the user's movements. The system collects and records the user's autonomous movement data over several consecutive movement cycles, which constitutes the user's natural movement pattern in this modality. Based on this collected autonomous movement data, the system generates a personalized assistance strategy representing the user's "flat-ground walking" habit through a series of calculations, and stores it in association with the user ID and the "flat-ground walking" modality identifier.

[0066] After the personalized assistance strategy is generated and saved, the system automatically switches to assistance mode. At this point, the user continues walking, and the system executes step S101 in real-time to obtain the user's current real-time movement state, and then executes step S105 to evaluate the degree of matching between this real-time movement state and the newly generated personalized assistance strategy. This degree of matching is a quantitative indicator that reflects the degree to which the user's current action conforms to the baseline pattern he / she has established.

[0067] Subsequently, in step S107, the system compares the calculated matching degree with a preset safety threshold. If the matching degree is lower than the threshold, for example, if the user suddenly stumbles or attempts to make movements unrelated to walking, the system will determine it as an abnormal state and immediately switch the exoskeleton's assist output to a preset safety mode to ensure user safety. If the matching degree is not lower than the safety threshold, it indicates that the user is walking stably according to their inherent pattern, and the process proceeds to step S108.

[0068] In step S108, the system calculates the most suitable magnitude and direction of assistance at the current moment based on the established personalized assistance strategy, and generates an assistance control signal to drive the exoskeleton's motors to output the corresponding auxiliary torque. In this way, the user can experience a smooth and natural assistance that is highly synchronized with their own movement rhythm and habits. Through this entire process, this embodiment achieves the technical effect of the exoskeleton automatically learning the user's movement characteristics and providing personalized and safe assistance without any pre-setting or manual adjustment.

[0069] Building upon the above embodiments, a further embodiment of this work further defines the specific sources of the sensor data. The control device acquires kinematic or physiological parameters from sensors mounted at different locations on the hip exoskeleton via its input / output interface. For example, the control device can receive data in real time from an inertial measurement unit (IMU) mounted on the thigh linkage, as well as data from a flexible plantar pressure sensor array integrated into the user's insole, via a controller area network (CAN) bus. In this way, the system obtains rich, multi-source input information, providing a solid data foundation for subsequent precise control.

[0070] In another progressive embodiment, the types of the kinematic or physiological parameters are more specifically defined. An inertial measurement unit (IMU) can provide triaxial acceleration and triaxial angular velocity data. The control device can use this data, through integration or attitude calculation algorithms, to accurately calculate the attitude angle of the thigh link relative to the vertical direction and the angular velocity of the hip joint. A plantar pressure sensor array can provide pressure distribution information for various areas of the user's foot. By analyzing the total pressure or pressure in specific areas, the ground contact and takeoff phases in the gait cycle can be accurately determined. These specific parameters provide key features with a high signal-to-noise ratio for subsequent motion mode recognition and state assessment.

[0071] Please see Figure 2 This document illustrates a specific embodiment of a motion modality recognition method. Building upon the previous embodiment, this embodiment further defines specific rules for identifying the user's current motion modality. First, in step S201, the system determines whether a swing leg identifier already exists. If not, it proceeds to step S202, where the swing leg is determined by calculating the changes in the hip joint angles of the left and right legs. For example, the leg with a positive hip flexion (forward swing) angular velocity is identified as the swing leg. After determining the swing leg, the system continuously monitors the plantar pressure of that swing leg in step S203. When the total pressure exceeds a threshold (e.g., 15% of the user's body weight), the system determines that a landing event has occurred.

[0072] Upon determining the moment of landing, the system immediately executes step S204 to check the plantar pressure of the supporting leg. If the pressure of the supporting leg is less than a very low threshold (e.g., 5% of the user's body weight), it indicates a brief period of airborne movement, and the system identifies the current motion mode as "running." If the supporting leg pressure does not meet this condition, the process proceeds to step S205, where the system obtains the hip joint angle of the swinging leg at the moment of landing. If this angle is greater than a preset threshold (e.g., 40 degrees), it indicates that the user lifted their leg high to climb stairs, and the system therefore identifies the mode as "climbing stairs." If the climbing stairs condition is not met, step S206 further determines the hip joint angle of the supporting leg at the moment of landing to distinguish between "walking on flat ground" and "climbing stairs." Through this set of clear rules based on biomechanical characteristics, rapid and accurate classification of several common motion modes is achieved.

[0073] In another progressive embodiment, the learning mode is specifically defined. The preset non-assisted learning mode is explicitly defined as a zero-assistance mode. In this mode, the control system continuously sends zero-torque commands to the motor driver. This puts the motor in a highly compliant following state, and its output end (connected to the exoskeleton linkage) can be easily driven by the user. The motor itself only overcomes internal frictional torque without applying any active assistance or resistance to the user's movement. This design ensures that the hip joint motion trajectory data collected during the learning phase represents the user's most realistic and natural movement performance, thereby guaranteeing the accuracy of the personalized assistance strategy.

[0074] In another progressive embodiment, the personalized assistance strategy is concretized as a motion reference curve. The process of generating this curve includes: after identifying a new motion mode, the system continuously acquires hip joint angle and angular velocity data sequences for 3 to 5 complete gait cycles in zero-assist mode. After acquisition, the system uses a B-spline curve fitting method to process these discrete data points, generating a smooth curve that characterizes the continuous changes in hip joint angle, angular velocity, and even angular acceleration with the percentage of gait cycles (0% to 100%). This B-spline curve is the motion reference curve for that mode, storing the user's personalized motion pattern in a compact mathematical form.

[0075] To further improve the quality of the motion reference curve, a preferred embodiment defines the data processing steps. Before performing B-spline fitting, the system first applies a Kalman filter to the acquired raw autonomous motion data sequence. The Kalman filter can effectively filter out the measurement noise inherent in the sensor itself, as well as the high-frequency physiological jitter that is unavoidable during user movement. The filtered data sequence is smoother and better reflects the essential trend of motion. The motion reference curve fitted based on this smooth data will have more stable and reasonable first derivatives (angular velocity) and second derivatives (angular acceleration), which is crucial for the subsequent generation of smooth, impact-free assist torque.

[0076] In another progressive embodiment, the method for determining the preset safety threshold is described in detail. After generating the motion reference curve, the system performs a backtracking calculation: for each raw data point (including angle and angular velocity) collected during the learning phase, the weighted distance between it and the corresponding phase point on the motion reference curve is calculated. Then, the system iterates through all data points and finds the maximum value among these weighted distances, denoted as D. max Finally, the preset safety threshold for this mode is set to 1.5*D. max In this way, the safety boundary is adaptively generated based on the statistical distribution of the user's own motion data, which is more scientific and personalized than using a fixed threshold.

[0077] In another progressive embodiment, a preset safety mode is explicitly defined. When the weighted distance between the current motion state and the reference curve, calculated in real time, exceeds the aforementioned adaptively set safety threshold during the assistance process, the controller immediately forces the torque command value sent to the motor driver to 0. This preset safety mode, which sets the assistance output torque to zero, means that the exoskeleton will instantly "release" when it detects an anomaly, without applying any interfering force to the user. This prevents the danger from being exacerbated by incorrect assistance when the user may lose balance or make evasive maneuvers, thus maximizing the user's safety.

[0078] In the last progressive embodiment, the form of the assist control signal is defined as a torque control signal.

[0079] To more comprehensively demonstrate the preferred implementation of this embodiment, a complete embodiment incorporating all the above-described technical features is described below. This embodiment simulates a complete scenario where a user wearing a hip exoskeleton transitions from walking on flat ground to climbing stairs.

[0080] After the user puts on the device, they begin walking on flat ground. Upon power-up, the system first acquires kinematic or physiological parameters using the IMU and plantar pressure sensors. When the hip joint angular velocity exceeds the activation threshold and the plantar pressure exhibits periodic changes, the system determines that the user has begun exercising. Subsequently, the system utilizes... Figure 2 The logic shown performs motion mode recognition and determines that the current mode is "walking on flat ground". System check found that a personalized assistance strategy for this user's "walking on flat ground" has not yet been established.

[0081] The system then automatically enters zero-assist mode, with the controller sending a zero-torque command to the motor driver. The user walks with a natural gait, and the system collects autonomous motion data for 3 to 5 gait cycles in the background. After collection, the system first applies a Kalman filter to smooth the raw data to remove noise. Then, using a B-spline curve fitting method, a motion reference curve representing the user's flat-ground walking pattern is generated based on the smoothed data. This curve stores the reference angle, angular velocity, and angular acceleration corresponding to any phase within the 0-100% gait cycle. Simultaneously, the system calculates the maximum weighted distance D between the learned data and the generated curve. max The safety threshold for this mode is set to 1.5 * D. max .

[0082] After learning is complete, the system seamlessly switches to assisted mode. While the user continues walking on flat ground, the system, within each control cycle: 1) acquires the current gait phase; 2) queries the reference angle, reference angular velocity, and reference angular acceleration corresponding to that phase from the "flat ground walking" reference curve; 3) acquires the current angle and angular velocity fed back by the sensors; 4) calculates the weighted distance between the current state and the reference curve, and confirms that it is less than 1.5D. max 5) If safe, calculate the target assist torque; 6) Send the torque control signal to the motor driver to provide assistance.

[0083] When a user approaches the stairs and begins to climb, their gait changes significantly. The system uses modal recognition logic ( Figure 2 The system detected this change: when the swing leg landed (with plantar pressure exceeding 15% of body weight), the system found that the hip angle of the swing leg at the moment of landing exceeded 40 degrees. Based on this, the system switched the movement mode from "walking on flat ground" to "climbing stairs". The system queried again and found that there was no personalized assistance strategy for the "climbing stairs" mode.

[0084] At this point, the control process automatically repeats the learning process: the system re-enters the preset non-assisted or low-intervention learning mode (zero-assisted mode in this example), prompting the user to climb a few steps normally. The system collects autonomous motion data for 3-5 climbing gaits, filters and smooths it, generates a new motion reference curve specific to the "climbing stairs" mode, and calculates the adaptive safety threshold for this mode. After learning is complete, the system switches back to assisted mode, and then provides precise assistance to the user's climbing motion using the same torque calculation formula based on this new "climbing stairs" reference curve. Throughout the process, safety monitoring is always online. If the user loses their footing while climbing the stairs, causing the deviation between the motion trajectory and the reference curve to momentarily exceed the threshold, the assistance torque will be immediately set to zero to ensure the user's safety.

[0085] This complete embodiment demonstrates how it seamlessly integrates motion modality recognition, online learning, personalized strategy generation, adaptive safety threshold setting, and feedforward and feedback-based torque control. Through the synergistic work of these technologies, the hip exoskeleton achieves high adaptability to different users and different movement scenarios, superior personalized experience, and reliable safety. Compared to existing fixed-mode control methods, it significantly improves the naturalness, comfort, and safety of human-computer interaction. For example, the matching degree between the assist curve and the user's natural gait can be improved by more than 30%, and the user does not need to perform any manual mode switching or cumbersome parameter adjustments.

[0086] Example 4 Those skilled in the art can understand this embodiment as a more specific description of Embodiment 1 and Embodiment 2.

[0087] like Figure 1 As shown, this embodiment provides an adaptive hip exoskeleton-assisted control method to ensure a better experience for different individuals. Based on each person's own motion change curve, a personalized motion assistance signal is generated. The specific steps are as follows: Step 1: Acquire sensor information, including but not limited to hip joint angle, angular velocity, IMU data, plantar pressure, video signals, and other sensor information.

[0088] Step two: Determine whether exercise has started based on sensor information. If exercise has not started, the assistance is zero. If exercise has started, proceed to step three.

[0089] Step 3: Determine the current motion mode based on sensor information. Proceed to Step 4.

[0090] Step four: Determine if a motion reference curve has been saved. If a motion reference curve exists, proceed to step five. If no motion reference curve exists, proceed to step six.

[0091] Step 5: Obtain the angle and angular velocity values ​​from the current feedback, and calculate the difference between the current point and the corresponding point on the motion reference curve. Proceed to Step 7 after completion.

[0092] Step Six: Keeping the motor assist at zero, record motion data for several cycles. The motion data mainly includes the hip joint angle and angular velocity. (Note that the motion mode identifiers should remain unchanged during these cycles; if they change, they should be cleared and re-recorded.) After recording, smooth the data to fit a smooth motion reference curve and record the current motion mode identifier. The fitted motion reference curve should provide values ​​for angle, angular velocity, and angular acceleration. Consider recording the maximum weighted distance between the fitted motion curve and the corresponding angle and angular velocity values ​​of the collected data points, as a reference for determining the threshold in Step Seven. End the entire process upon completion.

[0093] Step 7: Determine if the threshold is exceeded. If so, set the torque to zero and end the process; otherwise, proceed to step 8.

[0094] Step 8: Generate torque commands based on the motion reference curve under the current motion mode, send them to the motor, and end the entire process.

[0095] This embodiment first identifies the motion mode, and then determines whether to save the motion reference curve based on the identified motion mode. The motion curve is recorded without assistance to obtain more realistic human motion characteristics. The saved motion curve is smoothed to avoid abrupt changes in angle and angular velocity values. At the same time, the motion reference curve stores relevant information on angle, angular velocity, and angular acceleration. The position of the current point and the corresponding point on the motion reference curve is determined by a weighted combination of angle and angular velocity values, so that the final output torque is more in line with the actual human motion.

[0096] In step three, various methods can be used to determine the motion mode based on the combination of sensors, including but not limited to using established rules or neural networks to determine the current motion mode.

[0097] like Figure 2 As shown, in a specific scenario, a plantar pressure sensor combined with relevant information such as hip joint angle can be used to determine the current motion mode. The specific method is as follows: Check if there is a swing leg indicator. If yes, proceed to step three; otherwise, proceed to step two.

[0098] Calculate the hip joint angle changes for both legs. A clockwise rotation of the left hip joint is considered positive, and a counter-clockwise rotation of the right hip joint is considered positive. Label the leg with the positive hip joint angle change as the swing leg, and proceed to step three.

[0099] Calculate the changes in plantar pressure on the left and right legs, then proceed to step four.

[0100] Determine if the plantar pressure of the swinging leg has increased to exceed a threshold. If it has, the swinging leg is considered to have touched the ground, and proceed to step five. Otherwise, continue monitoring changes in plantar pressure.

[0101] Obtain the pressure information of the supporting leg at the moment the swinging leg lands, then proceed to step six.

[0102] Determine if the pressure value of the supporting leg is less than a given threshold. If it is less, mark the current state as running, proceed to step ten, and end the entire process. If it is greater, proceed to step seven.

[0103] Obtain the hip joint angle information on both sides at the moment the swing leg lands, then proceed to step eight.

[0104] Determine if the hip joint angle of the swinging leg at the moment of landing is greater than a given threshold. If it is, mark the current state as going upstairs, proceed to step ten, and end the entire process. Otherwise, proceed to step nine.

[0105] Determine if the hip angle of the supporting leg at the moment of landing is greater than a given threshold. If it is, mark the current state as walking on flat ground; otherwise, mark the current state as going downstairs. Proceed to step ten to end the entire process.

[0106] Switch the icon for the swing leg to the support leg.

[0107] The motion pattern determination algorithm proposed above is as follows: The pressure value of the swing leg's sole is used to determine whether the swing leg has landed, while the pressure value of the supporting leg is obtained at the same time. The human body's movement can be roughly divided into four main movement modes: walking on flat ground, going upstairs, going downstairs, and running. The pressure on the sole of the supporting leg at the moment the swinging leg lands is used to determine whether the current state is running. The angle of the swing leg's hip joint at the moment of landing is used to determine whether the current state is upstairs; Use the angle of the supporting leg's hip joint at the moment the swinging leg lands to determine whether the current state is going downstairs or walking on flat ground.

[0108] This invention constructs a personalized assist model by learning from the user's own real motion data, rather than relying on a general idealized curve. This allows the exoskeleton's assistance to be highly synchronized with the user's movement intentions and habits, significantly improving the accuracy, comfort, and safety of the assistance, and achieving adaptive control for different users and different sports scenarios.

[0109] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0110] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. An adaptive hip exoskeleton-assisted control method, characterized in that, Includes the following steps: Determine the user's initial state of motion, and after determining that the user is in the initial state of motion, identify the user's current motion mode; Determine whether a personalized assistance strategy for the current motion mode has been established; If the personalized assistance strategy is not established, then in the preset non-assisted learning mode, the user's autonomous movement data is collected, and the personalized assistance strategy is generated based on the autonomous movement data. If the personalized assistance strategy has been established, the user's current real-time motion status is obtained, and the degree of matching between the real-time motion status and the personalized assistance strategy is evaluated. When the matching degree is less than a preset safety threshold, the assist output of the hip exoskeleton will be switched to a preset safety mode. When the matching degree is greater than or equal to the preset safety threshold, an assistance control signal is generated based on the personalized assistance strategy.

2. The adaptive hip exoskeleton assistive control method according to claim 1, characterized in that, The degree of matching between the real-time motion state and the personalized assistance strategy is calculated using the following formula: ； Wherein, P is the matching degree, which measures the matching degree between the real-time motion state and the personalized assistance strategy, and takes the value [0, 1]; k is a custom allocation coefficient, and its value range is (0, 1). The motor reference angle value at one point on the reference trajectory given by the personalized assistance strategy; To and The corresponding real-time measured value of the motor angle returned by the current encoder; The reference angular velocity value of one point on the reference trajectory given by the personalized assistance strategy; To and The corresponding real-time measurement value of angular velocity returned by the current encoder; This is the maximum weighted distance value calculated when generating the reference trajectory.

3. The adaptive hip exoskeleton assistive control method according to claim 1, characterized in that, Acquire at least one state parameter that can characterize the user's lower limb movement state; The state parameters are kinematic or physiological parameters, and the user's initial state of movement is determined by the state parameters; The state parameters include at least one selected from hip joint angle, hip joint angular velocity, limb acceleration, plantar pressure signal, and electromyographic signal.

4. The adaptive hip exoskeleton assistive control method according to claim 3, characterized in that, The step of identifying the user's current motion mode includes: The landing event is determined by the plantar pressure signal of the swing leg, and the running mode is determined by combining the pressure signal of the supporting leg. The mode of going upstairs, going downstairs, or walking on flat ground is determined by the hip joint angle at the moment of the landing event.

5. The adaptive hip exoskeleton assistive control method according to claim 1, characterized in that, The preset non-assisted learning mode is the zero-assistance mode; The personalized assistance strategy is a motion reference curve; The step of generating the personalized assistance strategy includes: processing the autonomous motion data to fit and generate the motion reference curve; The processing of the autonomous motion data includes filtering or smoothing the autonomous motion data.

6. The adaptive hip exoskeleton assistive control method according to claim 5, characterized in that, When it is determined that a personalized assist strategy for the current motion mode has not been established, the current autonomous motion data is recorded in chronological order. The autonomous motion data includes angle data, angular velocity data, and torque value data. When it is determined that the recording of the autonomous motion data is complete, the recorded autonomous motion data is subjected to various filtering processes to obtain multiple corresponding reference motion curves; Multiple filtering processes employ any combination of the following filtering methods: Kalman filtering, moving average filtering, and first-order hysteresis filtering. The weighted distance between the autonomous motion data and the corresponding filtered data is calculated point by point to obtain the maximum value of the weighted distance. The filtered data sequence corresponding to the minimum value is selected from the results of multiple filtering methods as the motion reference curve. If, within a preset number of cycles, the number of times the matching degree between the real-time motion state and the personalized assistance strategy is less than a preset safety threshold exceeds a preset value, then the motion reference curve is automatically cleared to zero, and new motion reference curves are generated by re-collecting data.

7. The adaptive hip exoskeleton assistive control method according to claim 1, characterized in that, The determination of the preset safety threshold includes setting it based on the maximum value of the weighted distance between the autonomous motion data and the generated personalized assistance strategy; The formula for calculating the maximum value of the weighted distance is: ； in, The maximum value of the calculated weighted distance; k is a user-defined allocation coefficient, with a value range of (0, 1); The motor reference angle value at one point on the reference trajectory given by the personalized assistance strategy; To and The corresponding motor angle measurement value obtained from recording autonomous motion data; The personalized assistance strategy provides the reference angular velocity value of one point on the reference trajectory; To and The corresponding motor angular velocity measurement value obtained from recording autonomous motion data; The preset safety mode sets the assist output torque of the hip exoskeleton to zero.

8. The adaptive hip exoskeleton assistive control method according to claim 1, characterized in that, The assist control signal is a torque control signal, which is sent to the motor used to drive the hip joint exoskeleton; The formula for calculating the torque control signal is as follows: ； in, This is the torque control signal that is ultimately output to the motor; , , The coefficients are user-defined and can take values ​​from the entire real number field. , , The angle, angular velocity, and angular acceleration values ​​corresponding to the point on the motion reference curve that best matches the current real-time state in the proposed personalized assistance strategy. and This refers to the motor angle and angular velocity currently being acquired in real time.

9. The adaptive hip exoskeleton assistive control method according to claim 4, characterized in that, The step of identifying the user's current motion mode specifically includes the following steps: Step S1: Determine if a swing leg marker already exists. If yes, proceed to step S2. If no, calculate the changes in the hip joint angles of the left and right legs and mark the leg with a positive hip joint angle change as the swing leg. Step S2: Continuously calculate and monitor the plantar pressure value of the swing leg, and determine whether the plantar pressure value of the swing leg increases and exceeds the first threshold. If yes, proceed to step S3; otherwise, continue the calculation and monitoring. Step S3: Calculate the plantar pressure value of the supporting leg at the moment the swing leg lands, and determine whether the plantar pressure value of the supporting leg is less than the second threshold. If yes, mark the current state as running; otherwise, proceed to step S4. Step S4: Obtain the hip joint angles on both sides at the moment the swing leg lands, and determine whether the hip joint angle of the swing leg is greater than the third threshold. If yes, mark the current state as going upstairs; otherwise, proceed to step S5. Step S5: Determine whether the hip joint angle of the supporting leg is greater than the fourth threshold. If yes, mark the current state as walking on flat ground; otherwise, mark the current state as going downstairs.

10. An adaptive hip exoskeleton assistive control system, characterized in that, Perform the following procedure: Determine the user's initial state of motion, and after determining that the user is in the initial state of motion, identify the user's current motion mode; Determine whether a personalized assistance strategy for the current motion mode has been established; If the personalized assistance strategy is not established, then in the preset non-assisted learning mode, the user's autonomous movement data is collected, and the personalized assistance strategy is generated based on the autonomous movement data. If the personalized assistance strategy has been established, the user's current real-time motion status is obtained, and the degree of matching between the real-time motion status and the personalized assistance strategy is evaluated. When the matching degree is less than a preset safety threshold, the assist output of the hip exoskeleton will be switched to a preset safety mode. When the matching degree is greater than or equal to the preset safety threshold, an assistance control signal is generated based on the personalized assistance strategy.

Images